International Journal of Computer Theory and Engineering

Abstract

A generalized filter for enhancing fetal structures to aid feature extraction for design of automated clinical diagnosis and decision making system for fetal pathology identification has been presented in this paper. A new class of nonlinear filter, the combinational nonlinear mean median (CNLMM) filter with progressive switching median scheme for speckle detection and maximum likelihood mean estimator for speckle suppression has been proposed. Experimental analysis has been on a database of 1978 fetal images, obtained from high resolution scanners of various make with various features varying in shape, size and texture. For better analysis, the prominent fetal features have been grouped into linear, angular, curved and homogeneous features. A detailed comparison with various traditional filters has also been made and presented. It has been inferred from the quantitative analysis that the proposed CNLMM filter produces better and stable results compared to its counterparts for all the above feature structures, making it a suitable scheme for speckle suppression.